Marker points 3D matching based on Delaunay triangfulation structure and error dispersion

Abstract

Abstract Point cloud registration techniques based on marker points are widely used in optical 3D industrial measurements. However, in this process, marker points 3D matching methods are often haunted by low efficiency and accuracy. To improve the performance of marker points 3D matching, we propose a two-step method of ‘matching-verification’. In the matching process, Delaunay triangulation is introduced to extract the 3D structure of the marker points set, and then the 3D structure is deconstructed into 2D units for matching, which simplifies complexity and improves the efficiency of the algorithm. In the verification process, the mismatched pairs of points are located and removed by the method that is based on the error dispersion of initial matched results, and the initial transformation results are iteratively verified to obtain the optimal transformation matrix. The experimental results show that our method takes an average of 2.2 s for each matching, the average error of coarse registration point cloud is 0.075 mm and the root mean square is 0.219 mm, which effectively solves the problem of the low efficiency and accuracy of marker points 3D matching methods.